文章首先针对adarank, svm-map提出它们的共同点:均是优化IR measure的上界, 但是前者的界
是基于IR measure,
后者的界是基于最优排序和非最优排序的对,形式为:
仔细观察发现,将PERMU-rank写成类似(9)的形式,和SVMMAP的区别仅仅是把max变成了求和! 这篇文章太扯蛋了...
permurank利用了第二种bound的构造形式,说白了就是仿照SVM-MAP的loss来构造,构造出对第i个查询的loss形式为
对它进行优化.该形式中的B_i和C_i是最大违反约束集,实质是大PI_i*和大PI_i的子集.
这篇文章看后,有这么几点值得我借鉴的:
1, 从多篇类似的文章中,找到共通的框架, 利用这个框架,设计出不同的算法;
2, 在设计不同算法时,主要是从构造loss的方面考虑,其次是从使用的机器学习技术考虑(目前对我来说,哪种技术的实现源代码易用,benchmark效果好选哪种)
另外,svm-map的缺点是只针对MAP这种二元分类的IR-measure可行,因为在求delta-map时用到了这个observation.
- 大小: 2.9 KB
- 大小: 12.2 KB
- 大小: 17.3 KB
- 大小: 8.7 KB
- 大小: 12.2 KB
分享到:
相关推荐
标题中的“Analysis and optimization of distance measurement in MATLAB/SIMULINK model.pdf”揭示了文档的核心内容是关于在MATLAB/SIMULINK模型中对距离测量进行分析和优化的研究。这表明文档可能详细介绍了一个...
《优化条件风险价值》(Optimization of Conditional Value-at-Risk)是Rockafellar和Uryasev两位学者在2000年发表于《Journal of Risk》上的文章,该研究提出了一个新方法,用于降低金融投资组合的风险。文章的重点...
The objective of this framework is to go beyond the current limits of aircraft performance through simultaneous design optimization of aerodynamic shape, structural sizing and control system....
The concept of a measure of noncompactness is very useful for the mathematical community working in nonlinear analysis. Both these theories are especially useful in investigations connected with ...
This document describes techniques for optimizing (improving the speed of) computer programs written in C. It focuses on minimizing time spent by the CPU and gives sample source code transformations ...
A new approach to optimizing or hedging a portfolio of nancial instruments to reduce risk is presented and tested on applications. It focuses on minimizing Conditional Value-at-Risk (CVaR) rather ...
HOMER (Hybrid Optimization of Multiple Electric Renewables), the micropower optimization model, simplifies the task of evaluating designs of both off-grid and grid-connected power systems for a ...
1.5 Dynamics of Neural Learning in the Information-Theoretic Plane 1.6 Neural Nonlinear Activity in the Information-Theoretic Plane 1.7 Degree of Neural Complexity and Maximum Entropy 1.8 ...
The series in Vector Optimization contains publications in various fields of optimization with vector-valued objective functions, such as multiobjective optimization, multi criteria decision making, ...
3. ISO/IEC 15408:1999, "Information technology — Security techniques — Evaluation criteria for IT security." 4. Tsai, E. et al., "Efficient covert channel analysis via a revised information flow ...
Bayesian optimization is a powerful tool for the joint optimization of design choices that is gaining great popularity in recent years. It promises greater automation so as to increase both product ...
In this Letter the problem of optimization of speckle patterns in a ghost imaging (GI) system is addressed. The mutual coherence between the measuring matrix and the sparsifying dictionary matrix is ...
《Design and Optimization of Passive UHF RFID Systems》是一份关于超高频无源RFID系统设计与优化的宝贵资源,其中包含了多个关键领域的详细分析和技术探讨。RFID(Radio Frequency Identification)技术是现代...
Optimization of Power System Operation applies the latest applications of new technologies to power system operation and analysis, including several new and important areas that are not covered in ...
【优化条件风险价值(CVaR)】 在风险管理领域,条件风险价值(Conditional Value at Risk,简称CVaR)是一种衡量投资组合潜在损失的重要度量。它也被称为预期损失(Expected Shortfall)或平均最坏情况损失。...